Testing Language Ability
Speech-language pathologists use standardized tests to diagnose developmental language disorder (DLD). The tests should be valid, meaning they measure what they are supposed to. They should be reliable, meaning they do so consistently. But even when using highly valid and reliable tests, the speech-language pathologist must interpret the results carefully. One critical decision is what cut-off score to use. Many agencies continue to determine eligibility for speech and language services based on a score of 2 standard deviations below the mean (standard score of 70) on one test or a score of 1-1/2standard deviations below the mean (standard score of 77) on two tests. But not all tests of language ability are equally hard for people with DLD, so the same cut-off score will lead to different decisions on one test compared to another. For some tests, the proper dividing point between children known to beat risk for DLD and children who are at very low risk can be well above the levels used by agencies. This means that some children who need help might be overlooked because their test scores did not match a faulty standard.
Here, we need to get a bit technical because the stakes are high. When we speak of the proper dividing point, we are referring to the related concepts of “sensitivity” and “specificity.” Sensitivity refers to how well a test can accurately identify children who truly have a disorder. Specificity refers to how well a test accurately “passes” children who do not have the disorder. Both are important because we don’t want to set the bar at a point where we miss children who need help OR where we accidentally flag children whose language ability is fine. Any tests of language ability that are used to determine children’s eligibility for services should have acceptable levels of sensitivity and specificity and a clear dividing line– a “cutoff” score – below which children are considered at risk for DLD.
To determine the cut-score on a new test (known as the “index test”), one inspects the scores of many test-takers who have already been identified by a “reference standard” as being at risk for DLD or not. The reference standard might be a well-established test of language ability or the opinion of experts who are highly experienced in diagnosing DLD. One then determines which score on the new test leads to the best match with the risk status determined by the reference standard.
Even when the results of a test with good sensitivity and specificity suggest risk, additional steps should be taken to make sure that the diagnosis is accurate. A speech-language pathologist familiar with the child should apply their clinical judgment and experience. They should talk to parents, teachers, or other professionals who know the child to learn how the low language abilities are affecting the child with family, friends, or in the classroom. Information from these sources can make a difference in arguing for services. And such information will be indispensable if agencies continue to use tests with unknown sensitivity and specificity.